Last Friday, the White House ordered Anthropic to restrict the export of its AI models Fable and Mythos, citing unspecified national security concerns. Shortly after, Anthropic pulled both models from global access, making them unavailable for a week. This incident marks the first real test of whether the U.S. government can use export controls to contain frontier AI, similar to its past attempts with encryption and spyware. The outcome could influence how other AI labs navigate export regulations.
The ban followed two key events. First, Anthropic provided access to Mythos through its limited partner program to a South Korean telecom, which U.S. officials suspected had ties to China. Second, Amazon CEO Andy Jassy reportedly alerted the administration after his researchers found a way around Fable 5’s safeguards. Anthropic disputes the 'jailbreak' label, calling it a narrow, already-patched issue rather than a wholesale defeat of the model’s safety measures. The Commerce Department issued an export control directive, and Anthropic had to limit access to its products within roughly 90 minutes of being notified, according to some accounts.
Anthropic marketed Mythos since its April launch as a Doomsday cyber machine that could disrupt the internet if released widely. Before the ban, only around 150 vetted companies and government organizations had access to it. The company aimed to help defenders secure their software and services before malicious actors could reach similar capabilities. The episode highlights the challenges of using export controls to contain powerful cyber technologies, a strategy with a mixed historical record.
Source: techcrunch
